See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/257668768

A Phenomenological Approach to Modeling

Transport in Porous Media

Article in Transport in Porous Media April 2011

DOI: 10.1007/s11242-011-9926-3

CITATION READS

1 10

2 authors:

Jacob Bear Leonid Fel

Technion - Israel Institute of Technology Technion - Israel Institute of Technology
45 PUBLICATIONS 2,205 CITATIONS 64 PUBLICATIONS 389 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Leonid Fel on 02 March 2017.

The user has requested enhancement of the downloaded file.

Transp Porous Med
DOI 10.1007/s11242-011-9926-3

A Phenomenological Approach to Modeling Transport

in Porous Media

Jacob Bear Leonid G. Fel

Received: 19 September 2011 / Accepted: 10 December 2011

Springer Science+Business Media B.V. 2011

Abstract The objective of this article is to make use of the phenomenological approach
to construct models for the transport of extensive quantities, such as mass of a fluid phase,
mass of a component of a fluid phase, momentum of a phase and energy, in porous medium
domains. Special attention is devoted to express the fluxes of these extensive quantities,
especially the non-advective ones, as functions of their relevant driving forces, obeying the
principle of minimum entropy production. It is shown that for each extensive quantity, we
have a linear diffusive flux term, a non-linear diffusive term, and a dispersive flux term. The
latter is shown to be proportional to the velocity squared. In each case, the number of moduli
that describe fluid and porous matrix properties is determined. The momentum balance equa-
tion for a porous medium domain, which is the motion equation, is analyzed and simplified
for special cases, leading to Darcys law and to Brinkmans equation.

Keywords Phenomenological approach Transport in porous media Modeling Fluxes

Entropy production

List of symbols
c Concentration of -chemical species
D Molecular diffusivity
D Molecular diffusivity in a pm
E Extensive quantity (e.g., m, m , M, E )
E Energy
e Intensive quantity of E
F Force
E
f Rate of transfer of E from to , per unit volume of pm

J. Bear (B) L. G. Fel

Department of Civil and Environmental Engineering, Technion-IIT, Haifa, Israel
e-mail: cvrbear@techunix.technion.ac.il

J. Bear
Kinneret College on the Sea of Galilee, Zemach, Israel

123
 J. Bear, L. G. Fel

f Subscript for fluid

I Internal energy per unit mass
jE Microscopic flux of E (...per unit flu. area)
JE Macroscopic flux of E (...per unit flu. area)
JpmE Macroscopic flux of E (...per unit pm area)
k Permeability
M Momentum (=V)
m Mass
T Temperature
t Time
W = V + (V)T
VE Velocity of E
V Velocity (Vm )
z Vertical coordinate (pos. upward)
Subscript for -phase
Shear stress
E Source of E per unit mass
Superscript for -species
 Hydraulic radius
Volumetric fraction of -phase
Thermal conductivity
pm Thermal conductivity of pm
Fluid viscosity
Density
Stress

1 Introduction

The phenomenological approach considered here involves the construction of macroscopic

continuum models of phenomena in porous media on the basis of observable macroscopic
variables only, avoiding their derivation from microscopic equations. Its aim is to provide an
appropriate explanation of experimental observations of phenomena in the most simple way.
This approach also ensures that the developed models are thermodynamically correct.
Here, we use the phenomenological approach for the construction of mathematical models
that describe phenomena of transport of extensive quantities, such as mass, momentum and
energy of phases, and phase components in a porous medium (henceforth, pm) domain. The
latter, regarded as a single continuum, is composed of a solid matrix and a void space, with
both subdomains distributed all over the pm domain. The void space is occupied by one or
more fluids.
With no attempt to present a historical review, let us highlight a few points. Since Darcys
(1856) famous experiments in Dijon, that led to Darcys law, models of transport of extensive
quantities have been proposed and used. In earlier years, the focus was on single-phase mass
flow. In later years, and nowadays, the efforts involve modeling of multi-transport phenom-
ena, e.g., multiple phases, multiple (possibly) interacting chemical species, non-isothermal
conditions and deformable porous media, often simultaneously. Various methods have been
used to construct such models, written at the porous medium (=macroscopic) level, starting
from what happens at the pore (=microscopic) level. Homogenization and REV averaging

123
A Phenomenological Approach

are examples. Our objective in this article is to show how such models can be achieved by
the phenomenological approach.
The phenomenological approach is not new. In fact, Darcys law, obtained on the basis
of sand column experiments, is a phenomenological law. In the area of groundwater flow,
Dupuit (1863), Forchheimer (1901), and Boussinesq (1904a,b), and many others, followed
the same path. Richards (1931) extended Darcys to unsaturated flow. In petroleum engi-
neering, Muskat (1949) wrote the classical book on Physical Principles of Oil Production, in
which Darcys law was extended to two phase flow. Bear et al. (1968, Sect. 1.6) presented a
brief historical review on the development of the theory and applications of phenomena of
transport in porous media. Most of the developments were based on the phenomenological
approach. Approaches based on mixture theory, REV averaging and homogenization started
only around the a 1950s.
Even in earlier years, it was obvious that the core of any flow model is the balance equation
of the fluids mass. Accordingly, the development of any flow and transport model starts from
the general balance equation for a considered extensive quantity, E, of a fluid in a spatial
domain, U , bounded by a closed surface S (microscopic level of description):

Quantity of E Net quantity of E




Net production of E
accumulating entering U
= + in U . (1)

in U

in U

during t
during t during t

Written for an -phase that occupies a volumetric fraction (= U /U ) of a control vol-

ume U of a pm domain, and with the possibility that the considered E can cross any
interface, the above expression can be written in the form

Quantity of Net quantity of Net quantity of Net production

E accumulating E entering U E entering U

= + + of E in U (2)
in U through S through S during t
during t during t during t

When the last form of the balance equation is written for a small (e.g., parallelepiped)
volume in a pm domain around a point, and for a small time interval, and then letting the vol-
ume and the time interval shrink to zero, the equation takes the form of a partial differential
equation (PDE) that expresses the balance of E at a (i.e., any) point within the pm domain:
e
= J,tot
E
+ f
E
+  E , (3)
t
where e denotes E per unit volume of fluid -phase. In this equation, J,totE denotes the
total flux of E, with and in the moving -phase, per unit phase area, f denotes the rate
E

at which E is transferred into the -phase from all other (-)phases, per unit volume of pm,
and E is the rate of production of E per unit mass of the -phase. This balance equation,
obtained phenomenologically, describes the transport of any extensive quantity in an -phase
that occupies the entire void space or part of it, in a pm domain. Equation 3 is referred to as
the macroscopic level. Appropriate expressions have to be provided for the flux, the transfer
and the source terms. The term flux is used here to denote the quantity of E passing through
a unit phase area per unit time.
For a specific E and , this PDE is then written and solved for a specified pm domain
geometry and initial and boundary conditions.
We wish to emphasize that every term in (3), as in all balance equations in this article,
expresses what happens in the vicinity of a point in the pm domain (regarded as a continuum),
in terms of values of state variables, that may be regarded as average values for some small

123
 J. Bear, L. G. Fel

domain around the point, and coefficients that represent various aspects of the effects of the
solid matrix configuration in that small domain.
As stated earlier, the objective of this article is to write this equation for the specific cases
of mass of a fluid phase, mass of a component of a fluid phase, momentum of a fluid phase,
and energy of the pm as a whole, and, especially, to use the phenomenological approach to
express the various fluxes, transfer and source terms that appear in these equations in terms
of the observable variables of the transport problem. Here, variables are the density, the
concentration of a chemical species, the pressure, and the temperature, and measurements
or observations are implemented via appropriate instruments that provide the values of these
variable at a point, meaning an average of the vicinity of the point. This is certainly
true as the spatial variations of the values of these variables in the vicinity of any point can
be approximated as linear. Furthermore, any local change in pressure, concentration, and
temperature tends to propagate quickly in the immediate vicinity, so that the measurement
by an instrument provides a representative value at the point.

2 Specific Balance Equations

Based on the phenomenological macroscopic E-balance equation (3), following is a number

of specific cases.

2.1 Mass of a Fluid Phase

For the mass of an -fluid phase (density ) that occupies part the void space, at the vol-
umetric fraction , since there are no -mass sources within an -phase, i.e., m = 0, the
-fluid macroscopic mass balance equation takes the form

= A J,tot
m
+ f
m
, (4)
t
m
where J,tot denotes the total -mass flux (=mass per unit time per unit area of -phase), A
denotes the areal fraction of in a cross section, and f m denotes the rate of mass transfer
from all -phases to the -phase, e.g., by evaporation or condensation, per unit volume of
pm.
Note that we have made a distinction between the porosity, (=volume of void space
per unit volume of pm), and the areal porosity, A (=area of void in a planar cross section,
per unit area of cross section), and in the case of multiple phases between and A , with
A = A (1 , 2 , 3 ), in which 1 , 2 , 3 denote the components of the unit vector, , normal
to the considered cross section. Nevertheless, it is usually assumed (as we shall do below)
that A .

2.2 Mass of a Chemical Species in a Fluid Phase

We consider a chemical -species present in a fluid -phase, occupying part of the void
space. From (3), we obtain

m m
= A J,tot + f + m , (5)
t

m
in which J,tot denotes the total flux of mass of within , f
m denotes the rate of transfer
of m to the -phase from the solid, and from another fluid phase, if present in the void

123
A Phenomenological Approach

space, across a fluidfluid interface, e.g., by volatilization of , and m may include such
phenomena as radioactive decay or chemical reactions within the -phase, in which is
produced.

2.3 Momentum of a Fluid Phase

For the linear momentum, M, of a fluid phase ( f ) that occupies the entire void space, the
momentum balance equation takes the form:

V
= A Jtot
M
+ fsf
M
+  M , (6)
t
M denoting the flux of momentum (per unit -area), and f M denoting the rate at which
with Jtot sf
momentum is transferred from the solid phase to the fluid, across their common interface,
per unit volume of pm. In the above equation, the source of momentum,  M , is due to body
(and other) forces (per unit mass of the phase). Note that the velocity, V, has the meaning of
momentum per unit mass of the phase.
If we are interested in the momentum balance equation for the pm as a whole, say in
single phase flow, we write one equation for the fluid and one for the solid, and add them.
The momentum exchange terms, f M , will then vanish. Actually, in the phenomenological
approach, we can write the momentum balance equation directly for the pm as a whole.

2.4 Energy of a Fluid Phase

For E representing the energy of a fluid that occupies the entire void space, e = (I + 21 V 2 ),
i.e., the sum of the internal energy and the kinetic energy of the fluid, per unit fluid volume,
but not potential energy, the energy balance equation is

   E E
I + 21 V 2 = A Jtot + f sf +  E , (7)
t
E denotes the total energy flux carried in and by the fluid, by advection diffusion
n which A Jtot
(heat conduction) and dispersion (per unit area of pm), f sf E denotes the energy (=heat)
transferred from the solid to the fluid across their common interface, per unit volume of pm,
and  E denotes the source of energy within the fluid (per unit volume of pm and per unit
time).
Note that the potential energy is not included as we assume that gravity is the only body
force.
Since we are interested in the energy balance equation for the pm as a whole, we can write
(7) once for the fluid (f) and once for the solid (s) and sum up the two equations. The total
interphase energy transfer term will vanish, and we shall obtain:

   E
 
I + 21 V2 = Jpm,tot + E , Jpm,tot
E
= E
A J,tot . (8)
t
(f,s) (f,s) (f,s)

All (macroscopic) balance equations in this section have been obtained strictly by phe-
nomenological considerations, without resorting to any averaging technique. They contain
fluxes, rates of interphase transfers and sources of the considered Es, which will now be
discussed in detail.

123
 J. Bear, L. G. Fel

3 Fluxes

E , is defined as j E = eV E . This
At the microscopic level of description, the total flux of E, jtot tot
flux can be decomposed:

E
jtot = eV E = eV + e(V E V) = jadv E
+ jdif
E
, V = V ,
( )

i.e., the sum of an advective flux carrying E at the fluids (mass weighted) velocity, V( Vm ),
and a diffusive flux, resulting from molecular motion, associated with the difference (V E V).

3.1 Advective Fluxes

At the macroscopic level, with V( Vf ) denoting the fluids average mass weighted veloc-
ity, the advective fluxes in a pm domain (per unit area of fluid phase), associated with the
movement of the fluid, but not with any other driving force, are:

3.1.1 Mass of a Fluid Phase

e = , Jadv
m = V. We may regard V as a momentum per unit mass. Then it is a state variable,

like and and we need to include the momentum balance equation in the set of equations
to be solved for these variables. Another option is to regard Jm ( V) as a flux driven by
the force p. Then, we use the generic terms in the polynomial representation of Jadv
m

m
Jadv,i = ai j j p + bi jk i p k p + , (9)
where the coefficients are functions of the fluid properties , and of the geometry of the
void space. For sufficiently small p, only the first term on the r.h.s. is significant, leading
to the linear Darcys law. For higher values of p, the second term is added and we obtain
an expression which is analogous to the Dupuit (1863)Forchheimer (1930) non-linear flux
law (actually, they wrote p = aV + b(V)2 ). The option of V as a momentum density is
considered in this article.

3.1.2 Mass of a Chemical Species in a Fluid Phase

e = , Jadv
m
= V.

3.1.3 Momentum of a Fluid Phase

e = V, Jadv
M
= VV.

3.1.4 Energy of a pm

Assuming that only the fluid phase in motion, we have:

E

Jadv = I + 21 V 2 V.

3.2 Diffusive Fluxes

For the macroscopic diffusive fluxes, we have:

123
A Phenomenological Approach

3.2.1 Mass of a Fluid Phase

m = (V V) 0, i.e., no diffusion of the total mass.

e = , Jdif

3.2.2 Mass of a Chemical Species in a Fluid Phase

e = . In analogy to Ficks law in a fluid, and with a driving force, which is the gradient of
(average) species concentration, expressed by the mass fraction of the species, ( /),
the macroscopic diffusive flux is Jdifm = D , in which D
pm pm denotes the macro-
scopic coefficient of molecular diffusion. It is a second rank tensor that can be expressed,
for example, as the product of the scalar molecular diffusivity in a fluid continuum, and a
void space geometrical property, T, often called tortuosity, which is a second rank tensor,
Dpm = TD

3.2.3 Momentum of a Fluid Phase

M = V(VM V) = , in which ( p) denotes the stress in the fluid,

e = V, Jdif
is the shear or deviatoric stress in the fluid phase, p is pressure and denotes the unit
tensor.
In the momentum balance equation (6), we can express the first term on the r.h.s. in the
form
A Jadv+dif
M
A (VV ) + A p,
in which the A expresses the diffusive flux of momentum. since only is contributing to
the dissipation of energy.

3.2.4 Energy of a Fluid Phase

E

ef = f (If + 21 f Vf2 ), Jf,dif = f If + 21 Vf2 (VfE Vf ). At the microscopic level, i.e., in
a phase continuum, the diffusive ( conductive) energy flux is expressed by Fouriers law,
jEf,dif eE (VfE V) = T , in which T denotes the temperature, and denotes the
thermal conductivity of the phase. By the phenomenological approach, at the macroscopic
E
level, in the fluid, the diffusive energy flux, driven by T takes the form Jf,dif = f T ,

in which the f denotes the macroscopic thermal conductivity within the fluid phase inside
the void space.
E
For the pm as a whole, the diffusive flux of energy is expressed as Jpm,dif = pm T , in
which pm denotes the thermal conductivity for the pm as a whole, i.e., through the composite
material composed of the solid matrix and the fluid occupying the void space. Thus, pm
depends on both f and s and on the configuration of the two phases in the pm domain.
Altogether, we have three driving forces: for the diffusive mass of a species, V +
(V)T (see below) for the diffusive flux of momentum and T for the diffusive flux of
energy. We have not taken into account coupled processes (in the Onsager sense) (e.g., De
Groot and Mazur 1962, p. 30).
Note that in this article, we do not deal with coupled processes (in the Onsager sense).

3.3 Dispersive Fluxes

Around the 1950s, mainly in dealing with the quality of groundwater in aquifers, as associated
with dissolved chemical species, it was noted (e.g., summary in Bear 1972, Sect. 10.3), both

123
 J. Bear, L. G. Fel

in field experiments and in laboratory experiments, that a dissolved species is transported in a

pm domain, both in the general direction of the flow and also normal to it, in a way that could
not be explained merely by the movement of the fluid at its average velocity and by molecular
diffusion. This phenomenon was called dispersion. To bridge the discrepancy, an additional
flux was introduced at the macroscopic levelthe dispersive flux. It became obvious that
this additional flux is not a real flux, like the two fluxesadvective and diffusivediscussed
above; it does not exist at the microscopic level. It is a flux that has to be added in order to
compensate for the fact that the advective transport of the solute at the macroscopic level
is described in terms of the volume averaged velocity (and this was done because Darcys
law provides this velocity), while, actually, within the void space, the solute is transported at
every (microscopic) point by the local microscopic velocity. All this is in addition to molec-
ular diffusion, which is also included in the macroscopic description. Altogether, the total
macroscopic flux of a solute is expressed as the sum:

Jm = Jadv
m
+ Jdis
m
+ Jdif
m
. (10)
The dispersive flux will be considered below.
Although (10) is written for E = m , it is applicable for any E.

3.3.1 Dispersive Flux of a Solute

Over the years, research, using a variety of models, has led to various expressions that describe
the dispersive flux of a solute; not much research efforts have been devoted to the dispersion
of momentum and energy, although, in principle, this kind of flux should be present in the
respective (macroscopic) balance equation models.
The term Fickian expression has often been used, because of its resemblance to Ficks
law for molecular diffusion, i.e., a flux proportional to the gradient of concentration. In recent
years, some researchers, on the basis of field observations of plume development in aquifers,
have proposed various non-Fickian alternative approaches to the determination of plume
shape.
In what follows, we shall employ the phenomenological approach to suggest an expression
for the dispersive flux, based on the observation that this flux must depend on the velocity
(no dispersion if V = 0) and on the driving force (no dispersion if = 0).
Although the discussion in this section has been devoted to the dispersive flux of a solute,
the same phenomenon of dispersive flux is present also in the cases of transport of energy
and of momentum.

3.4 Non-Advective Fluxes

Within the framework of the phenomenological approach to modeling transport in pm, we

shall now derive expressions for the fluxes of extensive quantities, other than the advective
fluxes; these are the diffusive and dispersive fluxes mentioned above. We shall start with the
flux of a solute. The same approach will then be applied to the fluxes of other extensive
quantities.

3.4.1 Non-Advective Solute Flux

We make use of the generic terms in the polynomial representation of the total (macroscopic)
flux of a -species in a fluid phase. The total non-advective flux of , J , is produced by
(1) the velocity of the fluid phase, V, and (2) the driving force represented by ; the

123
A Phenomenological Approach

two are independent of each other. The flux is, thus, a smooth function of these two, factors,
J = J (, V, ).
Developing this functional relationship in a power series, up to third-order terms, we
obtain:

Ji (, , V) = Aik k + Bikl k l + Cikl Vl k + Diklm k l m
+ E iklm Vk l m + G iklm Vl Vm k , (11)
in which the various coefficients (Aik , Bikl , etc.) include the effect of .As everywhere else,
the Einstein summation convention is employed (e.g., Aik k (k) Aik k ). The
various A, B, . . . Gs are tensorial coefficients that are associated with fluid and pm proper-
ties, i.e, the geometry of the void space configuration; in multiphase flow these coefficients
depend also on fluid saturation.
In thermodynamics, the rate of entropy production, S , is related to the thermodynamic
driving force, X, and to the thermodynamic flux, Y by S = Yi X i (De Groot and Mazur
1962, p. 65). Furthermore, the rate of entropy production must be positive, i.e., S 0. Here,
the solute flux, J is driven by , which acts as a driving force. Thus, in this case,
X = , and its conjugate flux is Y = J . Altogether, we have
S (, , V) = Aik i k + Bikl i k l + Cikl i k Vl
+ Diklm i k l m + E iklm i k l Vm
+ G iklm i k Vl Vm . (12)
Note that we have extended the construction of S , as proposed De Groot and Mazur (1962),
for a flux linearly proportional to a driving force, to the non-liner case considered here.
The requirement that S be positive definite, i.e., S 0, leaves in (12) only the quadratic
and the two quartic terms,
S (, , V) = Aik i k + Diklm i k l m
+ G iklm i k Vl Vm 0. (13)
In (13), we note that certain symmetries exist in the three tensors A, D, and G:
Aik = A1ki , Diklm = Dkilm = = Dilmk , G iklm = G kilm = G ikml = G lmik , (14)
where the tensor Diklm is invariant under every permutation of the full symmetric group S4 .
Thus, with (13), and since
S =
J , , (15)
in which
V1 , V2  denotes the scalar product of the vectors V1 and V2 , it follows that the
non-advective flux is expressed as

Jnonadv,i = Aik k Diklm k l m G iklm k Vl Vm . (16)
The first two terms on the r.h.s. of (16) do not involve the velocity. They describe diffusion.
The first term is actually the diffusive flux expressed by Ficks law, with Aik Dij . The
second term represents a non-linear, or non-Fickian diffusive flux. The last term expresses
the dispersive flux discussed above, with as the driving force, but with a proportion-
ality to V 2 . This may be still considered a Fickian law, but it is different from the Fickian
expression mentioned in Sect. 3.3, in that here the flux depends on V 2 . It may be interesting to
mention that in one of the earliest works on dispersion, the work of Taylor (1953), concerning
dispersion in a capillary tube, the dispersive flux was also proportional to V 2 .

123
 J. Bear, L. G. Fel

A detailed analysis of the tensors A, D, and G in a 3D pm domain with a prescribed sym-

metry (e.g., isotropic, axisymmetric) shows that the increase in symmetry towards isotropy
causes a reduction in the number of independent moduli associated with these tensorial coef-
ficients. For an isotropic pmhighest symmetrywe need four moduli for the description
of the three tensorial coefficients:
d
Aik = aik , Diklm = (ik lm + il km + im kl ) ,
3
g2
G iklm = g1 ik lm + (il km + im kl ) , (17)
2
S (, , V) = a( )2 + d( )4 + g1 ( )2 V2 + g2
, V2 , (18)

with a, d, g1 , g2 > 0, and the non-advective flux presented above as (16) taking the form

Jnonadv,i ( , V) = ai d( )2 i g1 V2 i + g2
, VVi .
(19)

Again, in this equation, the first term expresses linear diffusion, the second term expresses
non-linear diffusion, and the last term expresses the dispersive flux of . In (19), we note the
driving force, and the dependence on V 2 .
For axisymmetric pm domains, with the axis of symmetry indicated by the unit vector e,
we need eleven (2 + 3 + 6) moduli to describe these three tensorial coefficients

S (, , V, e) = a1 ( )2 + a2
, e2 + d1 ( )4 + 2d2 ( )2
, e2
+ d3
, e4 + g1 V2 ( )2 + g2
, V2 + g3 V2
, e2
+ g4
e, V2 ( )2 + g5
e, V
V, 
, e
+ g6
e, V2
e, 2 , (20)

with the following thermodynamic constraints imposed on the eleven moduli:

a1 , a2 > 0, d1 , d3 > 0, d1 d3 > (d2 )2 ,

and, similar to the results obtained in Fel and Bear (2010),

g1 > 0, g1 + g2 > 0, g1 + g3 > 0, g1 + g4 > 0, g1 + g2 + g3 + g4 + g5 + g6 > 0,

(g1 )2 + g1 (3g2 + g3 + g4 + g5 + g6 ) + 2g2 (g3 + g4 + g6 ) > 21 (g5 )2 .

The non-advective momentum flux is:

Jnonadv,i (, , V, e) = a1 i + a2
, eei + d1 ( )2 i

+ d2 ( )2 ei +
, e2 i + d3
, e3 ei
+ g1 V2 i + g2
, eVi + g3 V2
, eei

+ g4
e, V2 i + g5
e, V
V, ei +
, eVi
+ g6
e, V2
e, ei , (21)

in which the first two terms describe the linear diffusive flux, with the tensorial coefficient of
diffusion depending on two scalar moduli, a1 and a2 . The next three terms, with moduli d1 ,
d2 , and d3 , describe the non-linear diffusive flux components. The remaining terms describe
the dispersive flux. We note that for this axially symmetric pm, the dispersivity coefficient is
defined by six dispersivity moduli, g1 , . . . , g6 .

123
A Phenomenological Approach

3.4.2 Non-Advective Momentum Flux

At the microscopic level, i.e., in a fluid continuum, the deviatoric stress, i j , which expresses
the dissipative part of the diffusive momentum flux, is related to the driving force Wi j (
i V j + j Vi ), which is a symmetric second rank tensor. With this in mind, at the macro-
scopic level, the non-advective flux of momentum depends on the fluids velocity, V, and on
a driving force, W, by the general constitutive relation
i j = i j (W, V) = Mi jkl Wkl + Ni jklpstr Wkl W ps Wtr + L i jklps Wkl V p Vs , (22)
where Mi jkl = Mi jkl = M jikl = Mi jlk , Ni jklpstr = N jiklpstr = . . . = Ntri jklps and
L i jklps = L jiklps = L i jlkps = L kli j ps = L i jklsp are tensorial coefficients that depend on
fluid and void-space properties.
The first term on the r.h.s. represents the linear diffusive flux of momentum, with the
fourth rank tensor Mi jkl standing for the usual fluid viscosity for a Newtonian fluid. The sec-
ond term, with the eighth rank tensorial coefficient, Ni jklpstr , is responsible for non-linear
viscous effects. The dispersive part of the non-advective momentum flux involves the sixth
rank tensorial coefficient L i jklps .
The corresponding rate of entropy production is:
S(W, V) = i j Wi j = Mi jkl Wi j Wkl + Ni jklpstr Wi j Wkl W ps Wtr + L i jklps Wi j Wkl V p Vs .
(23)
To facilitate the presentation of (23) for an isotropic pm, we introduce the following
notations for operations with the W-tensor,

W,  Wii ,
W, W Wi j W ji ,
W, W, W Wi j W jk Wki ,

W, W, W, W Wi j W jk Wkl Wli ,
WV, WV Wi j V j Wik Vk . (24)
With this notation, we get eleven (2 + 5 + 4) viscous moduli,
S(W, V) = M1
W, W + M2
W, 2 + N1
W, W2 + N2
W, W, W, W
+ N3
W, W, W
W,  + N4
W, W
W, 2 + N5
W, 4 + L 1
W, WV2
+ L 2
WV, WV + L 3
V, W, V
W,  + L 4
W, 2 V2 . (25)
When
W,  = 0, equivalent to V = 0, i.e., isochoric flow, we are left for the isotropic
case, with only five (1 + 2 + 2) moduli:
S(W, V) = M1
W, W + N1
W, W2
+ N2
W, W, W, W + L 1
W, WV2 + L 2
WV, WV. (26)
The corresponding non-advective momentum flux is

j = M1 Wi j + N1
W, WWi j + N2 Wik Wkl Wl j + L 1 V Wi j + L 2 Wik Vk V j .(27)
M 2
Jnonadv,i
The first term expresses the diffusive flux of fluid momentum. The next two terms express
the non-linear momentum flux. The last two terms express the dispersive flux of momentum
(proportional to V 2 ). For a Newtonian fluid, M1 , i.e., the fluids viscosity.

3.4.3 Non-Advective Heat Flux

We consider the entire pm, i.e., a rigid stationary solid matrix and a void space occupied by a
single fluid, with both solid and fluid phases at thermal equilibrium, i.e., a single temperature

123
 J. Bear, L. G. Fel

T . The development of the expressions for the non-advective fluxes of heat is similar to
those of solute, except that in this case, the diffusive flux has to take into account the heat
transported in both fluid and solid phases. Altogether (16) through (21) are valid, except that
the numerical values of the various coefficients are different, and the thermal diffusivity of
the pm depends on the porosity, and on the thermal conductivities of the two phases, but not
on their densities.
For example, for an isotropic pm, we may express the non-advective heat flux in the form
(16), replacing by T , Aik by pm,ik and omitting the non-linear diffusive term. We obtain:

H
Jnonadv,i = pm,ik k T G iklm
H
Vl Vm k T, (28)

i.e., the sum of a diffusive term and a dispersive one.

4 Interphase Transfers

E
Here, we consider the rate of transfer, f , of an extensive quantity, E, to an -phase from
all other -phases, including the solid, within a pm domain.

4.1 Transfer of Mass of a -Species of a Phase

Such transfer across the interface between adjacent phases in the pm domain may be due to
adsorption/desorption of , ion exchange, or solid dissolution. We shall not elaborate on this
issue as it is well known and will detract attention from the main issue of this article.
M
4.2 Transfer of Momentum of a Phase, ffs

We assume that (1) the fluid in the void space sticks to the solid surface (no-slip condition),
and (2) the solid may be in motion (e.g., due to deformation). The M-transfer, across the
solidfluid interface, is due to the fluids velocity gradient at the solid microscopic surface,
integrated over the entire fluidsolid interface area, per unit volume of porous medium. We
assume that the fluid is Newtonian, and that at a point in the porous medium domain, this
transfer is (1) proportional to the difference in average velocity between the fluid and the
solid matrix, (2) inversely proportional to some length characterizing the distance between
these two subdomains, e.g., the length characterizing the size of a pore (we shall use the
hydraulic radius, , of the void space, equal to the volume of the void space divided by the
interfacial solid-fluid area), (3) proportional to the total interfacial area between the phases,
and (4) proportional to the fluids viscosity. Thus,
  V j, f V j,s
M
f sf = Ri j , (29)
i 2
in which Ri j , a second rank symmetric tensor, is coefficients associated with the geometry
of the void space. Note that the volume of void space per unit solid-void surface area is equal
to the hydraulic radius of the void space.

4.3 Transfer of Energy of a Phase

There is no interphase energy transfer as we have assumed that all phases are at the same
temperature.

123
A Phenomenological Approach

5 Sources of Extensive Quantities

5.1 Source of Mass of a Phase

There are no mass sources within a phase.

5.2 Source of Mass of a Component of a Phase

This source may be due to decay, or production of components due to chemical reactions.
We shall not elaborate on this kind of source here.

5.3 Source of Momentum of a Phase

We are considering sources only in the fluid (but per unit volume of pm). This source is due
to forces acting on the fluid. In the case considered here, we have two sources of momen-
tum per unit volume of porous medium. One is due to body forces, here due to gravity,
 M = gz. The other is due to the pressure gradient in the fluid , A p, where we
have taken into account that the fluid occupies only (the fluid) part of any cross-sectional
area through the pm.

5.4 Sources of Energy of a Phase

E
There are four sources of energy: (1) chem chemH , due to heat generated by chemical

reactions in the fluid phase (if such reactions exist), (2) V (F), due to the work, per unit
volume of pm, by the body force F (= gz), (3) {V A ( )}, resulting from the
work of the viscous (shear) stress in the fluid, per unit volume of pm, and (4) V ( p)
due to the work of the pressure, per unit volume of the porous medium.

6 Final Balance Equations

In this section, we shall insert the sources, the rates of transfer, and the expressions for fluxes
of E in the balance equations presented in Sect. 2, in order to obtain the specific balance
equations in terms of the problem variables:
 p, T, V, , = 1, 2, . . . , N . We have not
counted as a variable because = ( ) . Altogether we have N + 3 variables, and N + 3
equation: N + 2 balance equations and one constitutive relation, = ( p, T, ).
In this section, to simplify the presentation, we will make the assumption that = A .

6.1 Mass Balance for a Fluid Phase

Inserting the advective mass flux in (4), replacing by (= A ), i.e., single phase, we
obtain:

= (V) . (30)
t
6.2 Mass Balance for a Chemical Species

From (5), rewritten for a single fluid phase, we obtain

123
 J. Bear, L. G. Fel

 
m m m
= V + Jdif + Jdisp + f +  m , (31)
t
into which we can now insert appropriate expressions for the non-advective -fluxes that
appear in Sect. 3.4. As an example, we shall make use of (19) (for the isotropic case):

  
= V a 2d( )2 g1 V2 + g2
, VV
t
m
+ f +  m . (32)

We note that the total -flux is made up of an advective flux, a (linear) diffusive flux, with a
representing the (scalar) coefficient of diffusion in a pm, Dpm , a non-linear diffusive flux,

and a dispersive flux, which is proportional to the fluids velocity squared and depends on
two (scalar) coefficients that represent the effect of the void space geometry.

6.3 Momentum Balance

Based on the discussion up to this point, we may now write the momentum balance equation
(6) in the form
V
+ VV = G, G = Jdif
M
+ Jdisp
M
p gz + fsf
M
. (33)
t
By combining this equation with the mass balance equation (30), we obtain another form
of the macroscopic momentum balance equation
DV  
= Jdif
M
+ Jdisp
M
p gz + fsfM
(34)
Dt
or, in view of (29),
DV   R
= Jdif
M
+ Jdisp
M
( p + gz) 2 (Vf Vs ). (35)
Dt 
In what follows, we shall (1) neglect the non-linear diffusive flux of momentum, (2) neglect
the dispersive flux of momentum, (3) assume a Newtonian fluid, and (4) express the diffusive
flux of momentum by V, i.e., assuming isochoric flow at the microscopic level. The
momentum balance equation (35) then reduces to
DV R
= V ( p + gz) 2 (Vf Vs ). (36)
Dt 
Sometimes, different s are used in the first term and last terms on the r.h.s. of (36): The first
is referred to as turbulent viscosity, while the second is the molecular viscosity.
The above equation may be regarded as a generalized motion equation. Let us consider a
number of simplifications:

Case A

We assume that the magnitude of the velocity is such that the viscous force, resisting the
flow, due to the transfer of momentum at the fluidsolid interface is much larger than both
the inertial force and the viscous resistance to the flow, i.e.,
 DV   R     R 
       
    2 (Vf Vs ), and  V   2 (Vf Vs ).
Dt  

123
A Phenomenological Approach

Then, the momentum balance equation reduces to

k
(Vf Vs ) = ( p + gz) , k = 2 R T , (37)

which is Darcys law, with k denoting the permeability, Thus, Darcys law, which is usually
regarded as a flux equation, is nothing but a simplified form of the momentum balance equa-
tion. The product Vf is usually referred to as the specific discharge of the fluid.

Case B

We neglect the inertial effects, but we maintain the internal viscous friction in the fluid. In
this case,
 V 
 
   |V V|.
t
Then, (36) reduces to
R
V ( p + gz) (Vf Vs ) = 0, (38)
2
known as the Brinkman equation. In it, we note two viscous terms: one which is due to the
momentum transfer from the solid to the fluid, and the other which is due to a fluid velocity
gradient, unless the (macroscopic) velocity is uniform everywhere, i.e., V = const.. It is
possible that the viscosity, , is not the same in these two terms, with one in the first term on
the r.h.s. of (38) depending also on the geometry of the void space. The Brinkman equation
is usually employed when we deal with a saturated pm bounded by a body of free water.

Case C

When the local acceleration, V/ t , cannot be neglected, e.g., when flow starts, or in
oscillatory flow, the momentum balance equation takes the form
V R
= V ( p + gz) 2 (Vf Vs ). (39)
t 
6.4 Energy Balance

We start from the energy balance equation (7). For the total energy flux, in the case of an
isotropic pm, we make use of (28), i.e., taking into account only the linear diffusive heat flux,
and the Fickian dispersive flux. Energy per unit volume of pm is added by (1) the work
done on the fluid phase by external body forces, e.g., gravity (V F V gz),
(2) work done on the fluid by the stress within the fluid, composed of the work done by the
viscous forces ( (V A )) and by the pressure forces, ( (V A p)) ), and (3) heat
produced by chemical reactions within the fluid (= H ). We obtain:
     
I + 21 V 2 = A I + 21 V 2 V A Jdif H
A Jdisp
H
t
+ {V ( A )} {V ( A p)} V gz +  H , (40)
in which the diffusive and dispersive fluxes of heat are defined by (28), and the shear stress,
, can be expressed by (22). Note that there is no exchange of heat between solid and fluid,
as we have assumed that both are at the same T .

123
 J. Bear, L. G. Fel

7 Summary and Conclusions

We have demonstrated how the phenomenological approach can be employed for the con-
struction of mathematical models that describe the transport of mass, momentum, and
energy in porous medium domains. In this approach, the models are constructed on the
basis of assuming the phenomena that occur at the macroscopic level envisioned as a
continuum.
Several advantages can be mentioned: (1) there is no need for up-scaling (e.g., by averag-
ing or homogenization), as the models are written directly at the macroscopic level, and only
phenomena that are considered relevant are included, (2) the positive definitiveness of the
entropy production is ensured, (3) all flux expressions are tensorially correct, (4) we obtain
also non-linear diffusive flux expressions, and (5) we obtain expressions for the dispersive
fluxes which arise because the models are written in terms of averaged fluid velocity, while
the real advective transport is produced by the local (microscopic) velocities, which vary
from point to point within the void space. We also obtain the correct number of coefficients
that are related to void space geometry, with their correct number of moduli that represent
various features of void-space geometry and symmetry characteristics. The transport models
derived here have to be validated experimentally. The coefficients in these models have to
be obtained experimentally for any given pm, making use of some inverse technique.
Initial and boundary conditions can also be obtained phenomenologically, directly at the
macroscopic level, based on the understanding of phenomena that occur on the boundaries
(e.g., specified pressure or flux, or continuity of flux). Because each of the balance equations
describes what happens at (i.e., in the vicinity of) a point, they are applicable to heterogeneous
pm domains.
The same approach is also applicable to cases not included in this article, e.g., multiphase
flow, non-elastic deformable porous medium (in which Vs = 0), and non-Newtonian fluids.
We recall that in the phenomenological approach we assume that the constitutive relation-
ships of the solid matrix and the fluids that occupy the void space are assumed to have the
same shape as at the microscopic level, but with different coefficients.

Acknowledgments This research was supported partly by funds from the European Communitys Seventh
Framework Programme FP7/2007-2013 under Grant Agreement No. 227286 as part of the MUSTANG project,
and partly by the Kamea Fellowship program.

References

Bear, J.: Dynamics of Fluids in Porous Media. Elsevier, Paris (1972)

Bear, J., Irmay, S., Zaslavsky, D.: Physical Principles of Water Percolation and Seepage. UNE-
SCO, Paris (1968)
Boussinesq, J.: Recherches thoriques sur lecoulement des nappes deau infiltres dans les sol, et sur le dbit
des sources. J. Math. Pure Appl. 10, 578 (1904a)
Boussinesq, J.: Recherches thoriques sur lecoulement des nappes deau infiltres dans les sol, et sur le dbit
des sources. J. Math. Pure Appl. 10, 363394 (1904b)
Darcy, H.: Les Fontaines Publiques de la Ville de Dijon, p. 647. Dalmont, Paris, (1856)
De Groot, S.R., Mazur, P.: Non-Equilibrium Thermodynamics. pp. 510. North-Holland Publishing Com-
pany, Amsterdam (1962)
Dupuit, J.: tudes thoriques et pratiques sur le mouvement des eaux dans les cannaux dcouverts et travers
les terrains permables. 2nd edn. Dunod, Paris (1863)
Fel, L.G., Bear, J.: Dispersion and dispersivity tensors in saturated porous media with uniaxial symme-
try. Transp. Porous Media 85(1), 259268 (2010)
Forchheimer, P.: Wasserbewegung durch boden. Z. Ver. Deutsch. Ing. 45, 17821788, (1901)

123
A Phenomenological Approach

Forchheimer, P.H.: Hydraulik. 3rd edn. Leipzig, Berlin Teubnes 542 pp (1930)
Muskat, M.: Physical Principles of Oil Production. McGraw-Hill, New York (1949)
Richards, L.A.: Capillary conduction of liquids through porous medium. Physics 1, 318333 (1931)
Taylor, G.I.: Dispersion of solute matter in solvent flowing slowly though a tube. Proc. R. Soc.
A 219(1137), 186203 (1953)

123
View publication stats